Faculty of Science, The Chinese University of Hong Kong (CUHK) - Professor WONG Hoi Ying (Department of Statistics)

Faces of CUHK Science

Our Eminent Scientists @ CUHK:
Big Data and Statistical Learning for Portfolio Risk Management

Prof. WONG Hoi Ying

Department of Statistics


The financial transaction nowadays no longer relies solely on economic theory and also involves interdisciplinary professions. Implementing statistical theory to assess risk and decision-making investment strategy can help minimise the risk of investment loss and maximise the return. Professor WONG Hoi Ying, Associate Dean (Student Affairs) of the Faculty of Science and Professor from the Department of Statistics, has successfully adopted big data and statistical learning methodologies to develop tools that combine statistical analysis and decision-making. An international asset management firm has exploited this sophisticated tool to construct portfolio selection strategies.


Balancing Risk and Returns
A balanced investment strategy combines asset classes in a portfolio in an attempt to balance risk and returns. Many people may think that “buying low and selling high” is an ideal strategy to improve profit. However, in reality, the role of stability is much more significant in making investment decisions. One of the critical indicators of decision-making is the analysis of financial data. WONG pointed out that these data contain a lot of "statistical noise" and “instability”. Therefore, an effective "filter" would be necessary to screen useful information out of the noise, thus enhancing the reliability of the risk evaluation tools and making investments to gain financial stability.

For example, the transaction cost is one of the instabilities of financial data. The net return will reduce if the investors do not take the number of transactions into account. This instability on profit returns could be ameliorated through data analysis to minimise the investment risks and boost the returns -- a delicate balance between statistics and risk management.


Statistical Analysis Far Beyond Numbers
Most people may consider the format of data has to be numerical. Nowadays, with the advancement of data science, data could be in many forms, including images, text, or even audio. Generally speaking, the statistical analysis comprises the distributional properties of randomness and establishing the benchmarks for decision-making. The distributional properties are the evaluation of the commonality and extreme value of the data set. On the other hand, the decision-making is target-oriented and determined based on the distributional properties. For instance, the stock price is full of market noise which distorts the picture of the underlying trend. Portfolio selection strategies based on the statistical decision are essential to screen the right stocks. One of WONG’s research interests is using statistical learning and big data methods to analyse the randomness of stock price and refine the strategic investment portfolio. Statistics is an interdisciplinary profession. Apart from accurate data analysis, the focus of different subject areas also plays a crucial role in decision making.


Are Big Data and Artificial Intelligence (AI) 100% Reliable?
The conventional statistical operation involves the two procedures described before, first, the analysis of distributional properties from a massive amount of data, and then the decision-making based on the analysis result. A large amount of statistical noise in the first procedure, which is irrelevant to the second procedure, hinders the efficiency of obtaining the optimized decisions from the results.

Taking the human genome as an example, there are only a few specific gene mutations causing lung cancers out of 31 million mega-base pairs from the whole human genome. The amount of DNA conveys the statistics concept of "dimensionality". The high dimensionality of DNA elevates the use of “sparsity” of relevant genes in lung cancer. It is tremendously challenging to extract valuable data from a massive amount of statistical noise. This constraint inspires WONG to develop a set of innovative statistical learning methods to integrate the specific application goals into the data analysis procedure, thereby identifying the sparsity of data effectively and sending the data directly to decision-making. This method can substantially prune the error of data analysis and refine the application efficiency of decision-making.

Securities are also a high-dimensional data set. The number of global stocks exceeds 40,000, coupled with the high-frequency trading volume in milliseconds. Identifying the sparsity of target data has become the key to minimise investment risks. WONG’s research has been implemented in the development of algorithmic trading and risk management platform, which helps the integration of estimation and optimization procedure. Looking forward, WONG will continue to work on the integration of capital allocation into the existing decision-making framework to make the best strategy for constructing a cooperation investment portfolio.


A simplified version of the risk calculator for some selected popular derivatives was launched as an online open-access educational platform developed by Professor Hoi Ying WONG and Professor Tony SIT for practitioners and public to understand and appreciate statistical learning theory for portfolio risk management. 


Artificial intelligence (AI) tools require the feeding of a massive amount of data and countless human-centric, guided empirical learning processes to reach automation. Oodles of data feeding value the role of statistics in AI development. AI is a “sublimation” of data analysis simplifying the tedious process. However, AI is not as omnipotent as it was shown in sci-fi movies to supersede human hands. Using AlphaGo as an example, the AI framework which beat Ke Jie, the number one ranked player in the world at the time, whose “knowledge” was acquired by extensive training from human play.

Ultimately, AI should not be viewed as entirely trustworthy during this revamp of technology. Scientists have to assess the benefits and limits of AI. “There is a limitation for anything, even the cleverest person in this world. During the expansion of the improvement of AI, scientists have to leverage different theory and cognition models carefully”, added WONG. Therefore, a solid theoretical foundation is essential to troubleshoot AI errors effectively. The most basic concept of statistics, the "law of large numbers", means that the more trials in repeated experiments, the frequency of events tends to a more stable value. With the advancement of computer science, collecting large amounts of data has become more efficient, which exemplifies the law of large numbers to a greater extent. Undoubtedly, AI would enhance the innovation of statistical research, especially the decision-making for vast subject areas.


Encountering Educational Challenges with Profound Modesty
WONG has joined CUHK for more than 20 years. Apart from devoting to his research, WONG takes up the role of Associate Dean (Student Affairs) in 2015. His roles in the teaching and deanery team inspire him about the significance of an open channel of communication. Regardless of being an educator or research supervisor, a professor should be an exemplary listener for the students despite their immature behaviour. Every student is unique and has their own dreams. “Our students are remarkably brilliant. As a teacher at CUHK, we need to be more humble,” WONG laughed, “Adopting an open and modest attitude helps me teach and supervise postgraduate students and makes for a great experience for everyone involved. Students though often make mistakes, but still, it is a vital part of the learning process. A good teacher should appreciate students for their true perseverance.” 



Professor WONG’s outstanding contribution in teaching has been well recognised by the CUHK Vice-Chancellor's Exemplary Teaching Award (2015, 2020) and Faculty Exemplary Teaching Award (2006, 2009, 2011, 2015 & 2020).


It is hard to imagine that students were dissatisfied with WONG's teaching performance in his early career. As a statistician upholding the “law of large numbers”, WONG continually evaluates his performance as a teacher and communicates with his students. Years passed, he develops his distinctive teaching style. His outstanding contribution in teaching has been well recognised by the CUHK Vice-Chancellor's Exemplary Teaching Award and Faculty Exemplary Teaching Award several times.

Hong Kong has gone through significant social changes in recent years which influence university education. WONG reaffirms the uniqueness of every student with different ways to express themselves. Teachers should be more humble to develop empathy and understanding with the students. As a leader of student affairs, he is committed to bridging the gap between the Faculty and students. He believes his interdisciplinary background would facilitate the communication among every stakeholder of the Faculty to research common ground and achieve a win-win situation, which is also in line with the "Multi-objective optimization" in statistics decision-making.


Text, Editing: Angela HUNG